Knowledge-based analysis of microarray gene expression data by using support vector machines

We introduce a method of functionally classifying genes by using gene expression data from DNA microarray hybridization experiments. The method is based on the theory of support vector machines (SVMs). SVMs are considered a supervised computer learning method because they exploit prior knowledge of gene function to identify unknown genes of similar function from expression data. SVMs avoid several problems associated with unsupervised clustering methods, such as hierarchical clustering and self-organizing maps. SVMs have many mathematical features that make them attractive for gene expression analysis, including their flexibility in choosing a similarity function, sparseness of solution when dealing with large data sets, the ability to handle large feature spaces, and the ability to identify outliers. We test several SVMs that use different similarity metrics, as well as some other supervised learning methods, and find that the SVMs best identify sets of genes with a common function using expression data. Finally, we use SVMs to predict functional roles for uncharacterized yeast ORFs based on their expression data.

Dynamic modeling of gene expression data

We describe the time evolution of gene expression levels by using a time translational matrix to predict future expression levels of genes based on their expression levels at some initial time. We deduce the time translational matrix for previously published DNA microarray gene expression data sets by modeling them within a linear framework by using the characteristic modes obtained by singular value decomposition. The resulting time translation matrix provides a measure of the relationships among the modes and governs their time evolution. We show that a truncated matrix linking just a few modes is a good approximation of the full time translation matrix. This finding suggests that the number of essential connections among the genes is small.

Fundamental patterns underlying gene expression profiles: Simplicity from complexity

Analysis of previously published sets of DNA microarray gene expression data by singular value decomposition has uncovered underlying patterns or “characteristic modes” in their temporal profiles. These patterns contribute unequally to the structure of the expression profiles. Moreover, the essential features of a given set of expression profiles are captured using just a small number of characteristic modes. This leads to the striking conclusion that the transcriptional response of a genome is orchestrated in a few fundamental patterns of gene expression change. These patterns are both simple and robust, dominating the alterations in expression of genes throughout the genome. Moreover, the characteristic modes of gene expression change in response to environmental perturbations are similar in such distant organisms as yeast and human cells. This analysis reveals simple regularities in the seemingly complex transcriptional transitions of diverse cells to new states, and these provide insights into the operation of the underlying genetic networks.

Importance of replication in microarray gene expression studies: Statistical methods and evidence from repetitive cDNA hybridizations

We present statistical methods for analyzing replicated cDNA microarray expression data and report the results of a controlled experiment. The study was conducted to investigate inherent variability in gene expression data and the extent to which replication in an experiment produces more consistent and reliable findings. We introduce a statistical model to describe the probability that mRNA is contained in the target sample tissue, converted to probe, and ultimately detected on the slide. We also introduce a method to analyze the combined data from all replicates. Of the 288 genes considered in this controlled experiment, 32 would be expected to produce strong hybridization signals because of the known presence of repetitive sequences within them. Results based on individual replicates, however, show that there are 55, 36, and 58 highly expressed genes in replicates 1, 2, and 3, respectively. On the other hand, an analysis by using the combined data from all 3 replicates reveals that only 2 of the 288 genes are incorrectly classified as expressed. Our experiment shows that any single microarray output is subject to substantial variability. By pooling data from replicates, we can provide a more reliable analysis of gene expression data. Therefore, we conclude that designing experiments with replications will greatly reduce misclassification rates. We recommend that at least three replicates be used in designing experiments by using cDNA microarrays, particularly when gene expression data from single specimens are being analyzed.

Saccharomyces Genome Database provides tools to survey gene expression and functional analysis data

Upon the completion of the Saccharomyces cerevisiae genomic sequence in 1996 [Goffeau,A. et al. (1997) Nature, 387, 5], several creative and ambitious projects have been initiated to explore the functions of gene products or gene expression on a genome-wide scale. To help researchers take advantage of these projects, the Saccharomyces Genome Database (SGD) has created two new tools, Function Junction and Expression Connection. Together, the tools form a central resource for querying multiple large-scale analysis projects for data about individual genes. Function Junction provides information from diverse projects that shed light on the role a gene product plays in the cell, while Expression Connection delivers information produced by the ever-increasing number of microarray projects. WWW access to SGD is available at genome-www.stanford.edu/Saccharomyces/.

GenEST, a powerful bidirectional link between cDNA sequence data and gene expression profiles generated by cDNA-AFLP

The release of vast quantities of DNA sequence data by large-scale genome and expressed sequence tag (EST) projects underlines the necessity for the development of efficient and inexpensive ways to link sequence databases with temporal and spatial expression profiles. Here we demonstrate the power of linking cDNA sequence data (including EST sequences) with transcript profiles revealed by cDNA-AFLP, a highly reproducible differential display method based on restriction enzyme digests and selective amplification under high stringency conditions. We have developed a computer program (GenEST) that predicts the sizes of virtual transcript-derived fragments (TDFs) of in silico-digested cDNA sequences retrieved from databases. The vast majority of the resulting virtual TDFs could be traced back among the thousands of TDFs displayed on cDNA-AFLP gels. Sequencing of the corresponding bands excised from cDNA-AFLP gels revealed no inconsistencies. As a consequence, cDNA sequence databases can be screened very efficiently to identify genes with relevant expression profiles. The other way round, it is possible to switch from cDNA-AFLP gels to sequences in the databases. Using the restriction enzyme recognition sites, the primer extensions and the estimated TDF size as identifiers, the DNA sequence(s) corresponding to a TDF with an interesting expression pattern can be identified. In this paper we show examples in both directions by analyzing the plant parasitic nematode Globodera rostochiensis. Various novel pathogenicity factors were identified by combining ESTs from the infective stage juveniles with expression profiles of ∼4000 genes in five developmental stages produced by cDNA-AFLP.

Recursive partitioning for tumor classification with gene expression microarray data

Precise classification of tumors is critically important for cancer diagnosis and treatment. It is also a scientifically challenging task. Recently, efforts have been made to use gene expression profiles to improve the precision of classification, with limited success. Using a published data set for purposes of comparison, we introduce a methodology based on classification trees and demonstrate that it is significantly more accurate for discriminating among distinct colon cancer tissues than other statistical approaches used heretofore. In addition, competing classification trees are displayed, which suggest that different genes may coregulate colon cancers.

Regulation of T cell receptor (TCR) β gene expression by CD3 complex signaling in immature thymocytes: Implications for TCRβ allelic exclusion

During αβ thymocyte development, clonotype-independent CD3 complexes are expressed at the cell surface before the pre-T cell receptor (TCR). Signaling through clonotype-independent CD3 complexes is required for expression of rearranged TCRβ genes. On expression of a TCRβ polypeptide chain, the pre-TCR is assembled, and TCRβ locus allelic exclusion is established. We investigated the putative contribution of clonotype-independent CD3 complex signaling to TCRβ locus allelic exclusion in mice single-deficient or double-deficient for CD3ζ/η and/or p56lck. These mice display defects in the expression of endogenous TCRβ genes in immature thymocytes, proportional to the severity of CD3 complex malfunction. Exclusion of endogenous TCRβ VDJ (variable, diversity, joining) rearrangements by a functional TCRβ transgene was severely compromised in the single-deficient and double-deficient mutant mice. In contrast to wild-type mice, most of the CD25+ double-negative (DN) thymocytes of the mutant mice failed to express the TCRβ transgene, suggesting defective expression of the TCRβ transgene similar to endogenous TCRβ genes. In the mutant mice, a proportion of CD25+ DN thymocytes that failed to express the transgene expressed endogenous TCRβ polypeptide chains. Many double-positive cells of the mutant mice coexpressed endogenous and transgenic TCRβ chains or more than one endogenous TCRβ chain. The data suggest that signaling through clonotype-independent CD3 complexes may contribute to allelic exclusion of the TCRβ locus by inducing the expression of rearranged TCRβ genes in CD25+ DN thymocytes.

Hypotension and inflammatory cytokine gene expression triggered by factor Xa–nitric oxide signaling

The signaling pathway initiated by factor Xa on vascular endothelial cells was investigated. Factor Xa stimulated a 5- to 10-fold increased release of nitric oxide (NO) in a dose-dependent reaction (0.1–2.5 μg/ml) unaffected by the thrombin inhibitor hirudin but abolished by active site inhibitors, tick anticoagulant peptide, or Glu-Gly-Arg-chloromethyl ketone. In contrast, the homologous clotting protease factor IXa or another endothelial cell ligand, fibrinogen, was ineffective. A factor Xa inter-epidermal growth factor synthetic peptide L83FTRKL88(G) blocking ligand binding to effector cell protease receptor-1 inhibited NO release by factor Xa in a dose-dependent manner, whereas a control scrambled peptide KFTGRLL was ineffective. Catalytically active factor Xa induced hypotension in rats and vasorelaxation in the isolated rat mesentery, which was blocked by the NO synthase inhibitor l-NG-nitroarginine methyl ester (l-NAME) but not by d-NAME. Factor Xa/NO signaling also produced a dose-dependent endothelial cell release of interleukin 6 (range 0.55–3.1 ng/ml) in a reaction inhibited by l-NAME and by the inter-epidermal growth factor peptide Leu83–Leu88 but unaffected by hirudin. Maximal induction of interleukin 6 mRNA required a brief, 30-min stimulation with factor Xa, unaffected by subsequent addition of tissue factor pathway inhibitor. These data suggest that factor Xa-induced NO release modulates endothelial cell-dependent vasorelaxation and cytokine gene expression. This pathway requiring factor Xa binding to effector cell protease receptor-1 and a secondary step of ligand-dependent proteolysis may preserve an anti-thrombotic phenotype of endothelium but also trigger acute phase responses during activation of coagulation in vivo.

Feedback-inducible nuclear-receptor-driven reporter gene expression in transgenic mice

Understanding nuclear receptor signaling in vivo would be facilitated by an efficient methodology to determine where a nuclear receptor is active. Herein, we present a feedback-inducible expression system in transgenic mice to detect activated nuclear receptor effector proteins by using an inducible reporter gene. With this approach, reporter gene induction is not limited to a particular tissue, and, thus, this approach provides the opportunity for whole-animal screens. Furthermore, the effector and reporter genes are combined to generate a single strain of transgenic mice, which enables direct and rapid analysis of the offspring. The system was applied to localize sites where the retinoic acid receptor ligand-binding domain is activated in vivo. The results identify previously discovered sources of retinoids in the embryo and indicate the existence of previously undiscovered regions of retinoic acid receptor signaling in vivo. Notably, the feedback-inducible nuclear-receptor-driven assay, combined with an independent in vitro assay, provides evidence for a site of retinoid synthesis in the isthmic mesenchyme. These data illustrate the potential of feedback-inducible nuclear-receptor-driven analyses for assessing in vivo activation patterns of nuclear receptors and for analyzing pharmacological properties of natural and synthetic ligands of potential therapeutic value.

Temporal mapping of gene expression levels during the differentiation of individual primary hematopoietic cells

A hierarchical order of gene expression has been proposed to control developmental events in hematopoiesis, but direct demonstration of the temporal relationships between regulatory gene expression and differentiation has been difficult to achieve. We modified a single-cell PCR method to detect 2-fold changes in mRNA copies per cell (dynamic range, 250–250,000 copies/cell) and used it to sequentially quantitate gene expression levels as single primitive (CD34+,CD38−) progenitor cells underwent differentiation to become erythrocytes, granulocytes, or monocyte/macrophages. Markers of differentiation such as CD34 or cytokine receptor mRNAs and transcription factors associated with their regulation were assessed. All transcription factors tested were expressed in multipotent progenitors. During lineage-specific differentiation, however, distinct patterns of expression emerged. SCL, GATA-2, and GATA-1 expression sequentially extinguished during erythroid differentiation. PU.1, AML1B, and C/EBPα expression profiles and their relationship to cytokine receptor expression in maturing granulocytes could be distinguished from similar profiles in monocytic cells. These data characterize the dynamics of gene expression accompanying blood cell development and define a signature gene expression pattern for specific stages of hematopoietic differentiation.

Coordinate regulation of lipogenic gene expression by androgens: Evidence for a cascade mechanism involving sterol regulatory element binding proteins

To gain more insight into the molecular mechanisms by which androgens stimulate lipogenesis and induce a marked accumulation of neutral lipids in the human prostate cancer cell line LNCaP, we studied their impact on the expression of lipogenic enzymes. Northern blot analysis of the steady-state mRNA levels of seven different lipogenic enzymes revealed that androgens coordinately stimulate the expression of enzymes belonging to the two major lipogenic pathways: fatty acid synthesis and cholesterol synthesis. In view of the important role of the recently characterized sterol regulatory element binding proteins (SREBPs) in the coordinate induction of lipogenic genes, we examined whether the observed effects of androgens on lipogenic gene expression are mediated by these transcription factors. Our findings indicate that androgens stimulate the expression of SREBP transcripts and precursor proteins and enhance the nuclear content of the mature active form of the transcription factor. Moreover, by using the fatty acid synthase gene as an experimental paradigm we demonstrate that the presence of an SREBP-binding site is essential for its regulation by androgens. These data support the hypothesis that SREBPs are involved in the coordinate regulation of lipogenic gene expression by androgens and provide evidence for the existence of a cascade mechanism of androgen-regulated gene expression.

Dynamics of Gene Expression Revealed by Comparison of Serial Analysis of Gene Expression Transcript Profiles from Yeast Grown on Two Different Carbon SourcesD⃞

We describe a genome-wide characterization of mRNA transcript levels in yeast grown on the fatty acid oleate, determined using Serial Analysis of Gene Expression (SAGE). Comparison of this SAGE library with that reported for glucose grown cells revealed the dramatic adaptive response of yeast to a change in carbon source. A major fraction (>20%) of the 15,000 mRNA molecules in a yeast cell comprised differentially expressed transcripts, which were derived from only 2% of the total number of ∼6300 yeast genes. Most of the mRNAs that were differentially expressed code for enzymes or for other proteins participating in metabolism (e.g., metabolite transporters). In oleate-grown cells, this was exemplified by the huge increase of mRNAs encoding the peroxisomal β-oxidation enzymes required for degradation of fatty acids. The data provide evidence for the existence of redox shuttles across organellar membranes that involve peroxisomal, cytoplasmic, and mitochondrial enzymes. We also analyzed the mRNA profile of a mutant strain with deletions of the PIP2 and OAF1 genes, encoding transcription factors required for induction of genes encoding peroxisomal proteins. Induction of genes under the immediate control of these factors was abolished; other genes were up-regulated, indicating an adaptive response to the changed metabolism imposed by the genetic impairment. We describe a statistical method for analysis of data obtained by SAGE.

Integrin-mediated Adhesion of Endothelial Cells Induces JAK2 and STAT5A Activation: Role in the Control of c-fos Gene Expression

Integrin-mediated adhesion induces several signaling pathways leading to regulation of gene transcription, control of cell cycle entry and survival from apoptosis. Here we investigate the involvement of the Janus kinase (JAK)/signal transducers and activators of transcription (STAT) pathway in integrin-mediated signaling. Plating primary human endothelial cells from umbilical cord and the human endothelial cell line ECV304 on matrix proteins or on antibody to β1- or αv-integrin subunits induces transient tyrosine phosphorylation of JAK2 and STAT5A. Consistent with a role for the JAK/STAT pathway in regulation of gene transcription, adhesion to matrix proteins leads to the formation of STAT5A-containing complexes with the serum-inducible element of c-fos promoter. Stable expression of a dominant negative form of STAT5A in NIH3T3 cells reduces fibronectin-induced c-fos mRNA expression, indicating the involvement of STAT5A in integrin-mediated c-fos transcription. Thus these data present a new integrin-dependent signaling mechanism involving the JAK/STAT pathway in response to cell–matrix interaction.

Global analysis of gene expression in pulmonary fibrosis reveals distinct programs regulating lung inflammation and fibrosis

The molecular mechanisms of pulmonary fibrosis are poorly understood. We have used oligonucleotide arrays to analyze the gene expression programs that underlie pulmonary fibrosis in response to bleomycin, a drug that causes lung inflammation and fibrosis, in two strains of susceptible mice (129 and C57BL/6). We then compared the gene expression patterns in these mice with 129 mice carrying a null mutation in the epithelial-restricted integrin β6 subunit (β6−/−), which develop inflammation but are protected from pulmonary fibrosis. Cluster analysis identified two distinct groups of genes involved in the inflammatory and fibrotic responses. Analysis of gene expression at multiple time points after bleomycin administration revealed sequential induction of subsets of genes that characterize each response. The availability of this comprehensive data set should accelerate the development of more effective strategies for intervention at the various stages in the development of fibrotic diseases of the lungs and other organs.